A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 11, 17:30 UTC
[LMCache] Fix breakage due to new LMCache version (#30216)
Signed-off-by: Nick Hill <nhill@redhat.com>
Nick Hill committed
Dec 10, 2025 at 11:52 UTC
6ccb7baeb1a124ad9b6e87fe9bbd48ae40830869
2 files changed
+3
-3
requirements/kv_connectors.txt
+1
-1
@@ -1,2 +1,2 @@
1
-lmcache
1
+lmcache >= 0.3.10.post1
2
nixl >= 0.7.1 # Required for disaggregated prefill
vllm/distributed/kv_transfer/kv_connector/v1/lmcache_integration/vllm_v1_adapter.py
+2
-2
@@ -27,7 +27,7 @@ from lmcache.v1.lookup_client.lmcache_async_lookup_client import (
27
LMCacheAsyncLookupServer,
28
)
29
from lmcache.v1.offload_server.zmq_server import ZMQOffloadServer
30
-from lmcache.v1.plugin.plugin_launcher import PluginLauncher
30
+from lmcache.v1.plugin.runtime_plugin_launcher import RuntimePluginLauncher
31
32
from vllm.attention.backends.abstract import AttentionMetadata
33
from vllm.config import VllmConfig
@@ -683,7 +683,7 @@ class LMCacheConnectorV1Impl:
683
self.api_server = InternalAPIServer(self)
684
self.api_server.start()
685
# Launch plugins
686
- self.plugin_launcher = PluginLauncher(
686
+ self.plugin_launcher = RuntimePluginLauncher(
687
self.config,
688
role,
689
self.worker_count,